A novel fault diagnosis technique in wireless sensor networks

In sensor networks, performance and reliability depend on the fault tolerance scheme used in the system. With increased network size traditional fault tolerant techniques have proven inadequate. Further, identifying and isolating the fault is one of the key steps towards reliable network design. Towards this, we propose two new algorithms to detect and substitute faulty nodes at different levels in the network. In the proposed approach, the network is divided into zones which are having a master for each zone. Moreover, the masters of the zones are connected in a De Bruijn graph based network. When a fault occurs, the masters are checked, tested. After that, the sensor nodes in the suspected zone are tested. Our fault model assumes communication, processing and sensing faults caused by hardware failures in a node. We analyzed the performance of the first algorithm according to the number of messages it needs to diagnose faulty nodes. In addition, the performance of a 4-node De Bruijn graph was also studied by measuring the end-to-end delay. Finally, the performance of the second algorithm was studied by measuring the fault detection accuracy.